CN102346246B - The method of the pulse signal in perception cognitive radio system is carried out based on wavelet transformation - Google Patents

Abstract

The invention provides a kind of method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation, first it carry out wavelet transformation to the signal received by perception receiver, and the quadratic sum of wavelet conversion coefficient in small echo support set in particular dimensions is calculated at each time point place, be normalized to obtain the decision statistics of wavelet transformed domain Energy-aware with noise variance on this yardstick to this quadratic sum again, then the statistical nature of described decision statistics is analyzed, so that calculate described decision statistics between the area of observation coverage on maximal value, and then according to determine judging threshold, finally obtained decision statistics and determined judging threshold are compared whether there is porch signal in the signal judging to receive, this law computation complexity is lower, is convenient to apply in systems in practice.

Description

The method of the pulse signal in perception cognitive radio system is carried out based on wavelet transformation

Technical field

The present invention relates to the method for the pulse signal in a kind of perception cognitive radio system, particularly a kind of method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation.

Background technology

Cognitive radio (CR) is the effective means improving the availability of frequency spectrum.Because the availability of frequency spectrum of radar frequency band is very low, therefore receive very large concern with the cognitive radio system of radar system shared band and had a lot of research work.For cognitive radio system, in order to realize utilizing effective chance of radar frequency spectrum, need the effective blank frequency spectrum detected in radar frequency band, and harmful interference is not caused to radar system, and need detect radar signal in certain frequency range after to exit this frequency range as early as possible.Therefore, how effectively gordian technique is perception and the various radar signal in location.In reality, there is various radar signal, as continuous wave radar signal, pulsed radar signal and frequency modulated(FM) radar signal (chirpsignals).Different radar signal forms, needs to adopt different perception and detection method.Modal a kind of radar is the pulsed radar adopting burst sequence radar signal, and pulsed radar is mainly used in aviation and controls, weather forecast and marine navigation.

In order to realize coexisting of communication system and radar, perception with identify that faint radar pulse signal becomes an important task.Radar Signal Detection in a lot of document is all based on energy measuring, namely compares, to judge whether radar signal exists with a thresholding by being exported by energy detector.The advantage of energy detector does not need any prior imformation about naive user signal.But energy detector has the following weak point.First, energy detection algorithm cannot differentiate primary user's signal, secondary user's signal and interference.Secondly, the uncertainty of energy detector to noise is more responsive, and when low signal-to-noise ratio (SNR), Performance Ratio is poor.And radar pulse may occur in a kind of random mode, makes energy detection algorithm be difficult to be competent at.Therefore, effective detection & localization of random radar pulse signal is become to a difficulties in Radar Signal Detection.

Wavelet transformation (Wavelettransform) is the tool of local singularity and irregular structure (edge and width etc. as pulse) in detection signal.Document " M.Frisch; H.Messer; " Theuseofthewavelettransforminthedetectionofanunknowntran sientsignal; " IEEETransactionsonInformationTheory, vol.38, no.2, pp.892-897, March1992 " in utilize the momentary signal of Wavelet Detection the unknown; Document " AkiraOhsumi; HiroshiIjima; TomokiKuroishi; " Onlinedetectionofpulsesequenceinrandomnoiseusingawavelet; " IEEETransactionsonSignalProcessing, vol.47, no.9, pp.2526-2531, September1999 " in utilize random pulses string signal in Wavelet Detection Noise observation signal; Document " Z.Tian, G.B.Giannakis, " Awaveletapproachtowidebandspectrumsensingforcognitiverad ios, " inProc.Ofthel ^stinternationalConferenceonCognitiveRadio " in propose a kind of broader frequency spectrum hole detection method utilizing wavelet transformation; the power spectrum density (powerspectraldensity; PSD) that first the method utilizes FFT to estimate in broadband; then utilize the edge of wavelet transformation to different spectral region in power spectrum density (black, ash or white space) to carry out detection & localization effectively, this energy detector is equivalent to the energy detector in frequency domain.But, when low SNR, be difficult to because noise may destroy porch detect porch.Some documents also prove, wavelet transformation spatially exists very strong relevance at different scale.For each porch existed in input signal, its wavelet conversion coefficient will produce Local Extremum on continuous print metric space, and the wavelet transformation of noise then declines rapidly.And be understood that, the wavelet conversion coefficient of noise at different scale same translation parameters place spatially will on the occasion of or negative value, and the wavelet conversion coefficient of porch will be identical symbol at different scale same translation parameters place spatially.Based on these conclusions, some documents propose wavelet conversion coefficient on some continuous metric spaces to be multiplied, amassing for carrying out detection & localization to the random pulses in Gaussian noise of the multi-scale wavelet transformation coefficient obtained.The method can strengthen the peak amplitude of local extremum effectively, and effectively reduces noise.But except some special circumstances, the long-pending probability density function of multi-scale wavelet transformation coefficient is difficult to derive, thus constrains its practicality.In document " RymBesrour; ZiedLachiri; NoureddineEllouze; " UsingmultiscaleproductforECGcharacterization, " ResearchLettersinSignalProcessing, Volume2009; pp.1-5 ", the feature interpretation amassed for ECG of multi-scale wavelet transformation coefficient, wherein thresholding is determined according to root mean square (RMS) value of wavelet conversion coefficient in corresponding scale, but the thresholding chosen like this cannot meet CFAR requirement.Utilize in document " E.Elsehely; M.I.Sobhy; " Detectionofradartargetpulseinthepresenceofnoiseandjammin gsignalusingthemultiscalewavelettransform, " IEEEInternationalSymposiumonCircuitsandSystems, Orlando; Florida; USA, May30-June2,1999; vol.3, pp.536-539 " wavelet coefficient different scale spatially local extremum sum carry out detection & localization pulse signal.But when SNR is lower, be difficult to determine that Local Extremum is produced by noise or produced by required pulse signal, thus detection perform will be deteriorated.

From upper description, existing 2 class radar pulse detection algorithms----namely based on Energy-aware method with based on wavelet multi-scale product method, reliability and the practical application of these perception algorithms is all limited because of respective shortcoming.Therefore, in order to radar pulse signal be detected more reliably, improving efficiency and the precision of detection simultaneously, effective frequency spectrum sensing method must be studied.

And according to document " AkiraOhsumi, HiroshiIjima, TomokiKuroishi, " Onlinedetectionofpulsesequenceinrandomnoiseusingawavelet, " IEEETransactionsonSignalProcessing, vol.47, no.9, pp.2526-2531, September1999 " proof known: the wavelet conversion coefficient of edge of a pulse signal (regionofsupport in the support set of corresponding scale spatially wavelet filter, RoS) be all be then 0 on the occasion of (or negative value) in other positions, and the wavelet conversion coefficient of white Gaussian noise is still white Gaussian noise.And along with the increase of metric space, SNR becomes large.As shown in Figure 1, it is white Gaussian noise (σ ²=1) and the wavelet conversion coefficient of pulse signal, wherein, label is that the diagram of (a) represents white Gaussian noise (σ ²=1) and pulse signal; The diagram that label is (b) represents metric space s=2 ¹upper wavelet conversion coefficient; The diagram that label is (c) represents metric space s=2 ²upper wavelet conversion coefficient; The diagram that label is (d) represents metric space s=2 ³upper wavelet conversion coefficient.As can be seen here, porch wavelet conversion coefficient can be propagated on each yardstick, and the concentration of energy of wavelet conversion coefficient is in the support set of wavelet filter, and the wavelet conversion coefficient of noise does not have this feature.Therefore, the pulse signal how utilizing these features to carry out correct perception and locate in cognitive radio system, has become those skilled in the art's problem to be solved.

Summary of the invention

The object of the present invention is to provide a kind of method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation.

In order to achieve the above object and other objects, the method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation provided by the invention, comprise step: 1) wavelet transformation is carried out to the signal received by perception receiver, and calculate the quadratic sum of wavelet conversion coefficient in small echo support set in particular dimensions at each time point place, then be normalized to obtain the decision statistics of wavelet transformed domain Energy-aware with noise variance on this yardstick to this quadratic sum; 2) analyze the statistical nature of described decision statistics, so as to calculate described decision statistics between the area of observation coverage on maximal value, and then according to determine judging threshold, wherein, P _ffor false-alarm probability, its value presets, for the maximal value of described decision statistics, λ _efor judging threshold, H ₀represent in the signal received there is not porch signal, represent at H ₀under, probability; 3) obtained decision statistics and determined judging threshold are compared whether there is porch signal in the signal judging to receive.

In addition, the described method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation also can be included in region that decision statistics is not less than described judging threshold and search out the maximal value of decision statistics to estimate the step of the position of porch.

In sum, the method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation of the present invention is by the statistical study to decision statistics, judging threshold can be obtained, can realize thus carrying out efficient and reliably perception to radar pulse signal, and computation complexity is lower, be convenient to application in systems in practice.

Accompanying drawing explanation

Fig. 1 is white Gaussian noise (σ ²=1) and the wavelet conversion coefficient of pulse signal.

Fig. 2 is the schematic flow sheet carrying out the method for the pulse signal in perception cognitive radio system based on wavelet transformation of the present invention.

Fig. 3 is the schematic diagram of pulse signal s (n) to be detected.

Wavelet transformed domain porch energy detector T when Fig. 4 is SNR=8dB _{l, M}the testing result schematic diagram of (n).

Fig. 5 is for working as false-alarm probability P _fdifferent scale s=2 when=1% ^l(L=4,5) upper wavelet transformed domain porch energy detector with the performance curve of SNR.

Embodiment

Refer to Fig. 2, the method for the pulse signal come in perception cognitive radio system based on wavelet transformation of the present invention mainly comprises the following steps:

The first step: wavelet transformation is carried out to the signal received by perception receiver, and calculate the quadratic sum of wavelet conversion coefficient in small echo support set in particular dimensions at each time point place, then with noise variance on this yardstick, this quadratic sum is normalized to obtain the decision statistics of wavelet transformed domain Energy-aware.The present embodiment take signal as monopulse for example illustrates, but not as limit.Such as, suppose that the signal received by perception receiver can be expressed as: y (n)=As (n)+v (n) (1≤n≤N), wherein, A is pulse height, s (n) is pulse signal, v (n) is noise, carries out wavelet transformation to it, namely at yardstick s=2 ^lon wavelet transformation be: again according to calculate decision statistics T _{l, M}(n), wherein, M is integration sample number parameter, its by determine, for wavelet function original function.

Second step: the statistical nature analyzing described decision statistics, so as to calculate described decision statistics between the area of observation coverage on maximal value, that is: calculate definition of T _{l, M}n () maximal value between the area of observation coverage on [1, N] is: and then according to determine judging threshold, wherein, P _ffor false-alarm probability, its value presets, λ _efor judging threshold, H ₀represent in the signal received there is not porch signal, represent at H ₀under, probability, and for [1, N] between each the area of observation coverage, it will be a stochastic variable.Described false-alarm probability can preset the value of a regulation between 1%-10%.

Table 1 gives as observation sample length N=1000 and the false-alarm probability P pre-set _fwhen=1%, wavelet transformed domain decision statistics (porch energy detector) T when Different L and M _{l, M}the decision threshold λ of (n) _evalue, the result in table is added up according to 50000 independent experiment results and is obtained.

Table 1.

3rd step: obtained decision statistics and determined judging threshold are compared whether there is porch signal in the Noise random pulses observation sample signal judging to receive.For the false-alarm probability P pre-set _f, wavelet transformed domain porch energy detector T _{l, M}n () can choose appropriate threshold value λ from above table _e, then by T _{l, M}(n) and selected threshold value λ _ecompare to judge whether porch signal exists, and result be expressed as follows:

$T_{L,M}\left(n\right)=\left\{\begin{array}{cc}0,& T_{L,M}\left(n\right)<{\mathrm{\&lambda;}}_E\\ T_{L,M}\left(n\right),& T_{L,M}\left(n\right)\&GreaterEqual;{\mathrm{\&lambda;}}_E\end{array}\right.---\left(29\right)$

So [1, N] divide into the region be made up of " 0 " and the region be made up of " non-zero " between the area of observation coverage be made up of N number of sample, wherein each " non-zero " region shows have a porch signal to exist.Suppose there be D " non-zero " region (namely detecting D porch signal), and record the initial time sequence number B of m porch _mwith end time sequence number E _m, m=1,2 ..., D.Although the positional information of porch is unknown, T can be utilized _{l, M}n the maximal value of () correctly estimates the position of porch.So, in " non-zero " region corresponding to m porch, by search T _{l, M}n the maximal value of () estimates the positional information of m porch, can be expressed as follows:

${\hat{n}}_m=\arg \underset{B_m\&le;n\&le;E_m}{\max }\left|T_{L,M}\left(n\right)\right|,m=\mathrm{1,2},...,D.$

Be explained to the theory origin of method of the present invention below:

1, wavelet transformation

Small echo is defined as function ψ (t) that average is 0, and square integrable [8].Small echo ψ (t) is by contraction-expansion factor s=2 ^jstretching be expressed as

${\mathrm{\&psi;}}_{2^j}\left(t\right)=\frac{1}{2^j}\mathrm{\&psi;}\left(\frac{t}{2^j}\right)---\left(1\right)$

Function f (t) is at yardstick s=2 ^jcan convolution be expressed as with the wavelet transformation (WT) of position t:

$W_{2^j}f\left(t\right)=f*{\mathrm{\&psi;}}_{2^j}\left(t\right)---\left(2\right)$

The dyadic wavelet transform of the f (n) (1≤n≤N) of discrete signal can be expressed as sequence:

$\mathrm{Wf}={\left(W_{2^j}f\left(n\right)\right)}_{j\&Element;\{\mathrm{1,2},...,J\}}---\left(3\right)$

In formula, W is dyadic wavelet transform operator.About the detailed introduction of wavelet transformation, can list of references " MartinVetterli; CormacHerley; " Waveletandfilterbanks:theoryanddesign; " IEEETransactionsonSignalProcessing, vol.40, no.9, pp.2207-2232, September1992 ".

2, signal model

For simplifying the analysis, single pulse signal (directly can be generalized to the situation of multiple pulse signal) is considered:

s(n)＝[u(n-n ₀)-u(n-n ₁)]，1≤n≤N，(4)

In formula, n is sampling sequence number, and u () is unit step function, n ₀and n ₁for the start/stop time of pulse signal, the width of pulse is Δ D=(n ₁-n ₀).Then, suppose that the Noise Received signal strength at receiver place has following form:

y(n)＝A·s(n)+v(n)，1≤n≤N，(5)

In formula, A is the amplitude of pulse signal, and v (n) is additive white Gaussian noise (AWGN), and its average is 0, and variance is namely without loss of generality, suppose the SNR of pulse signal is defined as:

$\mathrm{SNR}=10\mathrm{lo}{g}_{10}(A^2/{\mathrm{\&sigma;}}_v^2)---\left(6\right)$

Pulse signal in detection formula (5) depends on its porch signal of extraction.

3. for the wavelet transformed domain porch energy detection algorithm of high-efficiency pulse detection & localization

3.1 based on the edge of a pulse energy detector of wavelet transformation (WT)

From the linear characteristic of wavelet transformation, the Dyadic Wavelet Transform of y (n) can be write as:

$z_j\left(n\right)=W_{2^j}y\left(n\right)=A{W}_{2^j}s\left(n\right)+W_{2^j}v\left(n\right),j=\mathrm{1,2},...,J---\left(7\right)$

More existing documents (as " S.Mallat; W.Hwang; " Singularitydetectionandprocessingwithwavelets; " IEEETransactionsonInformationTheory, vol.38, no.2, pp.617-643, March1992 " and " S.Mallat, S.Zhong, " Characterizationofsignalsfrommultiscaleedges, " IEEETrans.onPatternAnalysisandMachineIntelligence, vol.14, no.7, pp.710-732, July1992 ") in show be still random white Gauss noise, and variance can be expressed as:

${\mathrm{\&sigma;}}_{v,j}^2=E\left({\left|W_{2^j}v\left(n\right)\right|}^2\right)=\frac{{\left|\right|\mathrm{\&psi;}\left|\right|}^2{\mathrm{\&sigma;}}_v^2}{2^j}---\left(8\right)$

In addition, document " AkiraOhsumi; HiroshiIjima, TomokiKuroishi, " Onlinedetectionofpulsesequenceinrandomnoiseusingawavelet; " IEEETransactionsonSignalProcessing, vol.47, no.9, pp.2526-2531, September1999 " middle proof, the wavelet transformation of s (n) can be expressed as:

In formula, for original function, τ=τ (2 ^j) (> 0 makes for translation parameters supporting domain be (without loss of generality, suppose given small echo supporting domain be symmetrical).For Haar small echo (Haarwavelet), it is at yardstick s=2 ^jon supporting domain be [-τ, τ]=[-2 ^j-1, 2 ^j-1].When adjacent pulse edge is positioned at same wavelet filter h _jn in the supporting domain of (), then its wavelet conversion coefficient can produce mutual interference.Therefore, the out to out J of wavelet transformation need meet following condition:

2 ^J＜ΔD(10)

At yardstick s=2 ^lthe pointwise product of (1≤L≤J) upper wavelet conversion coefficient is defined as:

t _L(n)＝|z _L(n)| ²，1≤n≤N(11)

At yardstick s=2 ^lon (1≤L≤J), wavelet transformed domain porch energy detector is at h _lto t on the supporting domain [-τ, τ] of (n) _ln () carries out integration, then use obtain after the output of integrator is normalized " decision statistics " of wavelet transformed domain porch energy detector.So " decision statistics " of wavelet transformed domain porch energy detector is defined as following stochastic variable:

$T_{L,M}\left(n\right)=\frac{1}{{\mathrm{\&sigma;}}_{v,L}^2}\underset{k=-M+1}{\overset{+M}{\mathrm{\&Sigma;}}}{\left|z_L(n+k)\right|}^2---\left(12\right)$

In formula, 1≤M≤τ (2 ^l), for calculating T _{l, M}n the sample number of () is Δ L=2M.Then, by by " decision statistics " T _{l, M}n thresholding λ that () and one pre-sets _ecompare, to judge whether there is porch signal.Make H ₀for " 0 " is supposed, namely there is not porch signal; Make H ₁for " 1 " is supposed, namely there is porch signal.Therefore, in order to determine whether porch signal exists, Hypothesis Testing Problem can be expressed as:

H ₀：y(n)＝v(n)

H ₁：y(n)＝As(n)+v(n)(13)

So, at hypothesis H ₀under, T _{l, M}n center that () is degree of freedom 2M distribution.And at H ₁under, T _{l, M}n () is degree of freedom 2M and non-centrality parameter non-central distribution, wherein E _{l, M}n () equals:

$E_{L,M}\left(n\right)=\underset{k=-M+1}{\overset{+M}{\mathrm{\&Sigma;}}}{\left|{\mathrm{AW}}_{2^L}s(n+k)\right|}^2---\left(14\right)$

In formula, integration sample number parameter M can be chosen for and make detection probability maximum, chooses by following formula:

$M^{*}=\arg \underset{1\&le;M\&le;2^L}{\max }{E}_{L,M}\left(n_0\right)/\sqrt{4M}=\underset{k=-M+1}{\overset{+M}{\mathrm{\&Sigma;}}}{\left|{\mathrm{A\&Psi;}}_{2^L}\left(k\right)\right|}^2/\sqrt{4M}---\left(15\right)$

Based on above consideration, " decision statistics " that formula (12) defines is obeyed and is distributed as follows:

$T_{L,M}\left(n\right)=\left\{\begin{array}{cc}{\mathrm{\&chi;}}_{2M}^2\left(0\right)& H_0\\ {\mathrm{\&chi;}}_{2M}^2\left(\mathrm{\&eta;}\right)& H_1\end{array}\right.---\left(16\right)$

From formula (9), E _{l, M}n () is at n=n ₀and n=n ₁place obtains maximal value, and its maximal value can be expressed as:

$E_{L,M}\left(n_0\right)=E_{L,M}\left(n_1\right)=A^2\underset{k=-M+1}{\overset{+M}{\mathrm{\&Sigma;}}}{\left|{\mathrm{\&Psi;}}_{2^L}\left(k\right)\right|}^2---\left(17\right)$

At n=n ₀and n=n ₁place, the SNR of wavelet transformed domain porch energy detector reaches maximum, can be calculated as:

${\left(\mathrm{SNR}\right)}_{L,M}=10\mathrm{lo}{g}_{10}\frac{E_{L,M}\left(n_0\right)}{{\mathrm{\&sigma;}}_{v,L}^2}10\mathrm{lo}{g}_{10}[\frac{A^2}{{\mathrm{\&sigma;}}_v^2}\&CenterDot;\frac{2^L\&CenterDot;\underset{k=-M+1}{\overset{M}{\mathrm{\&Sigma;}}}{\left|{\mathrm{\&Psi;}}_{2^L}\left(k\right)\right|}^2}{{\left|\right|\mathrm{\&psi;}\left|\right|}^2}]---\left(18\right)$

About T _{l, M}the decision threshold λ of (n) _eby observation area [1, N] the upper false-alarm probability P arranged in advance _fdetermine.In fact, T _{l, M}the decision threshold λ of (n) _ecan at hypothesis H ₀lower calculated off-line arranges appropriate judging threshold with determining and can being stored in a look-up table as requested for detection algorithm.Assuming that P _fn () is T _{l, M}the false-alarm probability of (n), and equal for all n (1≤n≤N), then can obtain:

P _F(n)＝1-(1-P _F) ^1/N(19)

So, decision threshold λ _edetermine by following formula:

P _F(n)＝P(T _L，M(n)＞λ _E|H ₀)(20)

But, for common demand P _f≤ 5%, when observation sample number N is larger, P _fn () will be tending towards 1 very soon, now will corresponding to T _{l, M}the maximal value of (n).Therefore, in order to determine thresholding λ _e, definition of T _{l, M}n () maximal value between the area of observation coverage on [1, N] is:

$T_{L,M}^0=\underset{1\&le;n\&le;N}{\max }{T}_{L,M}\left(n\right)---\left(21\right)$

For [1, N] between each observation, it will be a stochastic variable.Now, false-alarm probability P _fmay be defined as:

$P_F=\mathrm{Pr}(T_{L,M}^0>{\mathrm{\&lambda;}}_E|H_0)---\left(22\right)$

In order to determine decision threshold λ _e, for each scale parameter L and the integration sample number parameter M that determined by formula (15), for enough large independent observation interval [1, N], calculate its maximal value respectively then, to about the false-alarm probability P pre-set _fthresholding carry out statistical estimate.Above-mentioned table 1 gives as observation sample length N=1000 and the false-alarm probability P pre-set _fwhen=1%, wavelet transformed domain porch energy detector T when Different L and M _{l, M}the decision threshold λ of (n) _evalue, the result in table is added up according to 50000 independent experiment results and is obtained.

In order to wavelet transformed domain porch energy detector T is described _l,Mn the validity of () and method of the present invention, provides simulation example below and is illustrated.First given pulse signal s (n) to be detected as shown in Figure 3, this signal comprises pulse signal and 2 impact signals that 4 have distinct pulse widths, wherein 4 pulses are positioned at interval [101,175], [301,450], [701,740] and [851,855], 2 impact signals are δ (n-600) and δ (n-900).

In detection, use Haar small echo, Wavelet-based Scale Space L=5, integral parameter M=8.Wavelet transformed domain porch energy detector T when Fig. 4 gives SNR=8dB _{l, M}the testing result of (n).Fig. 4 (a) is depicted as pulse signal s (n) to be detected, Fig. 4 (b) is depicted as pulse signal y (n)=As (the n)+v (n) containing white Gaussian noise, wherein A=2.5 during SNR=8dB.Fig. 4 (c), 4 (d) and 4 (e) sets forth yardstick L=3,4, the wavelet transformation of y (n) on 5.Fig. 4 (f) then gives wavelet transformed domain porch energy detector T _{l, M}the value of (n), wherein decision threshold λ _e=145.90 be according to table 1 arrange correspond to false-alarm probability P _fthe thresholding of=1%.As seen from Figure 4, pulsewidth Δ D > 2 ⁵the pulse signal of=32 can effectively detect.But the pulse signal of pulsewidth Δ D=5 < 32 can not detect effectively, and impact signal can be effectively suppressed.As can be seen here, simulation result and theoretical analysis completely the same.

In addition, wavelet transformed domain porch energy detector T is demonstrated _{l, M}the performance of (n), and with existing multi-scale product method P _{l, 2}n () [is defined as P _{l, 2}(n)=z _l(n) z _l-1(n)] carry out Performance comparision.In simulations, the test signal of use is unit step signal s (n)=u (n-300).By formula choose integral parameter M ^*, and decision threshold λ is set by table 1 _e.Fig. 5 gives as false-alarm probability P _fdifferent scale s=2 when=1% ^l(L=4,5) upper wavelet transformed domain porch energy detector with the performance curve of SNR.In order to compare, in figure, give existing multi-scale product method P _{l, 2}the simulation result of (n).As known in the figure, for different metric space s=2 ^l(L=4,5), wavelet transformed domain porch energy detector performance be better than multi-scale product method P _{l, 2}n the performance of () is about 2dB.

In sum, the method of the pulse signal come in perception cognitive radio system based on wavelet transformation of the present invention is different from existing Energy-aware method and multi-scale wavelet product method in the cognitive radio system that existing communication system and radar coexist, which overcoming existing Energy-aware method cannot the shortcoming of position pulse position and poor performance, and performance is better than existing multi-scale wavelet product method, can more reliable, effective perception radar pulse signal, and can determine the parameter such as position and pulse width of radar pulse, and computation complexity is lower.

Above-described embodiment only listing property illustrates principle of the present invention and effect, but not for limiting the present invention.Any person skilled in the art person all can without departing from the spirit and scope of the present invention, modify to above-described embodiment.Therefore, the scope of the present invention, should listed by claims.

Claims (4)

1. carry out a method for the pulse signal in perception cognitive radio system based on wavelet transformation, it is characterized in that comprising step:

1) wavelet transformation is carried out to the signal received by perception receiver, and calculate the quadratic sum of wavelet conversion coefficient in small echo support set in particular dimensions at each time point place, then be normalized to obtain the decision statistics of wavelet transformed domain Energy-aware with noise variance on this yardstick to this quadratic sum;

2) analyze the statistical nature of described decision statistics, so as to calculate described decision statistics between the area of observation coverage on maximal value, and then according to determine judging threshold, wherein, P _ffor false-alarm probability, its value presets, for the maximal value of described decision statistics, λ _efor judging threshold, H ₀represent in the signal received there is not porch signal, represent at H ₀under, probability;

3) obtained decision statistics and determined judging threshold are compared whether there is porch signal in the signal judging to receive.

2. the as claimed in claim 1 method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation, characterized by further comprising in the region that decision statistics is not less than described judging threshold, search out decision statistics maximal value to estimate the step of the position of porch.

3. the method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation as claimed in claim 1, it is characterized in that: when the signal received by perception receiver is expressed as: y (n)=As (n)+v (n) (1≤n≤N), it is at yardstick s=2 ^lon according to carry out wavelet transformation, and according to calculate decision statistics T _l,M(n), wherein, M is integration sample number parameter, its by determine, A is pulse height, and s (n) is pulse signal, and v (n) is noise, and L is scale parameter, for wavelet function original function, and 2 ^l< Δ D, Δ D are the distance value at adjacent pulse edge; Parameter n ₀the initial time of indicating impulse signal s (n).

4. the method carrying out the pulse signal in perception cognitive radio system based on wavelet transformation as claimed in claim 1, is characterized in that: described false-alarm probability is chosen in the scope of 1%-10%.